Challenge Details

Data and Protocol

The AI4PAIN 2026 dataset includes a total of 65 participants, carefully partitioned to ensure a robust evaluation framework. The AI4Pain dataset is divided into three subsets: training (41 participants), validation (12 participants), and testing (12 participants). As in previous editions of the challenge, the training and validation sets will be released early, while the test set will consist of unseen data to guarantee an unbiased evaluation of submitted models.   

This year’s challenge focuses on the analysis of physiological responses to painful stimulation using signals such as  Electrodermal Activity (EDA), Blood Volume Pulse (BVP), Respiratory (RESP), and Peripheral Oxygen Saturation (SpO₂). The task extends previous editions by introducing a problem that involves both pain detection and pain localisation, requiring participants to classify each sample into one of three categories: No Pain (NP), Hand Pain (HP), or Arm Pain (AP).  Labeled training and validation sets will be released in April 2026, while the unlabeled test set will be provided in May 2026. 

Participants have the flexibility to use any combination of available modalities, as there are no separate modality-specific challenges. They are encouraged to explore various feature extraction techniques, machine learning models, and deep learning architectures. However, they must strictly adhere to the predefined training, validation, and testing partitions, as test set labels will remain undisclosed. Although participants may report results using the training and validation sets, only the performance obtained on the test set will be used for the final ranking of the Grand Challenge. 

During the testing phase, participants will submit predicted labels for the test set, allowing organisers to compute and share classification results. Each team will be allowed up to five test label submissions. Additionally, participants are required to write and submit a paper detailing their methodologies, findings, and insights gained from the challenge, which will be presented at the PAAIn Workshop during the ACII 2026 conference in Puebla, Mexico. For more details, please refer to the Submission page.

Use of External Data: To ensure fairness and comparability, participants are strictly prohibited from using any external datasets for training, validation, or testing. The AI4PAIN 2026 dataset is the only permitted data source for model development. 

To register your team, go to the Challenge Registration page.